Re: [PATCH v4 03/25] io_uring/kbuf: add support for kernel-managed buffer rings

[Joanne Koong <joannelkoong@gmail.com>]

On Fri, Feb 6, 2026 at 5:40 AM Askar Safin <safinaskar@gmail.com> wrote:
>
> Joanne Koong <joannelkoong@gmail.com>:
> > Add support for kernel-managed buffer rings (kmbuf rings)
>
> Is it true that these kbufs solve same problem splice originally meant for?
> I. e. is it true that kbuf is modern uring-based replacement for splice?
>
> Linus said in 2006 in https://lore.kernel.org/all/Pine.LNX.4.64.0603300853190.27203@g5.osdl.org/ :
>
> > The pipe is just the standard in-kernel buffer between two arbitrary
> > points. Think of it as a scatter-gather list with a wait-queue. That's
> > what a pipe _is_. Trying to get rid of the pipe totally misses the
> > whole point of splice().
>
> So, kbuf is modern version of exactly this?

I don't think this is related to kmbufs. Zero-copying is done through
registered buffers (eg userspace registers sparse buffers for the ring
ahead of time and then on the kernel side, those sparse buffers can be
filled out by the kernel to point to relevant folios), which is
separate from kmbufs. In fuse, kmbufs are used for non-zero-copyable
payloads, and leveraged primarily as a way to minimize the memory
footprint (especially once incremental buffer consumption for kmbufs
is added, which I'm planning to do next).

Thanks,
Joanne

>
> --
> Askar Safin